Re: Register Allocation Algorithm..

[Jamie Lokier <egcs at tantalophile dot demon dot co dot uk>]

Daniel Berlin wrote:
> > Joe Buck wrote:
> > > Since the register allocator was written, we (the GCC project) have
> > > obtained permission from IBM and Rice University to use their register
> > > allocator software patents in GPL code (after complicated negotiations
> > > between FSF/IBM/Rice lawyers), which makes possible the new register
> > > allocator.
> > 
> > Is that permitted with GPL compilers other than GCC, or does the
> > permission only apply to GCC?
> 
> IIRC, GCC.
> But i don't quite remember what the letter said.
> > 
> > A slightly weaker question: can I be sued if I fork GCC and develop it
> > independently?
> You can be sued whether you do it or not.
> What exactly are you trying to ask?

I'm asking what the extent of the permission from IBM and Rice
University is, and who does it apply to?

Do we now have the situation where if I copy some code from GCC and use
it in another GPL'd project (e.g. for example a virtual machine, or fork
of GCC), I am not protected by the letter you mentioned?

In my mind, a whole line of development which is part of GCC, but which
is dangerous to use in another free software project, would be quite
contrary to ideals of Free Software.  Many people are lax about this
sort of thing but the FSF is often very careful.

Since I have not heard of this agreement with IBM and Rice before, and I
have no idea what the agreement says, and I'm interested in Free
compiler technology, I'd really appreciate knowing whether I can now --
as a result of the FSF/IBM/Rice University negotiations -- reasonably
assume it is ok to use graph colouring register allocation in a GPL'd
(or LGPL'd) program of my own.  And if not, is it ok to copy portions of
GCC's code?

I'm not asking the precise question of "Can I be sued?".

Merely, and I know it's a rather fuzzy question, "Can I assume it is 
_reasonably_ safe to use graph colouring techniques now?"

Just publishing the agreement that was reached might be a simple way to
answer this.

Thanks,
-- Jamie